Recognition and Reporting of AKI in Very Low Birth Weight Infants

Abstract

Background and objectives

AKI is associated with both increased short-term morbidity and mortality and greater long-term risk for CKD. This study determined the prevalence of AKI among very low birth weight infants using a modern study definition, evaluated the frequency of AKI diagnosis reporting in the discharge summary, and determined whether infants were referred to a pediatric nephrologist for AKI follow-up.

Design, setting, participants, & measurements

Records of very low birth weight infants admitted to a level IV neonatal intensive care unit from 2008 to 2011 were reviewed. AKI was classified using the Kidney Disease: Improving Global Outcomes definition modified to include only serum creatinine.

Results

AKI occurred in 39.8% of 455 infants; 75 (16.5%) infants experienced multiple episodes of AKI, and 8 (2%) infants were discharged with an abnormal last creatinine. Updated clinical risk index for babies score >10 (odds ratio, 12.9; 95% confidence interval, 7.8 to 21.4) and gestational age <28 weeks (odds ratio, 10.6; 95% confidence interval, 6.8 to 16.7) were strongly associated with AKI in univariate analyses. AKI was associated with increased mortality (odds ratio, 4.0; 95% confidence interval, 1.4 to 11.5) and length of stay (11.7 hospital days; 95% confidence interval, 5.1 to 18.4), even after accounting for gestational age, birth weight, and updated clinical risk index for babies score. AKI was recorded in the discharge summary for only 13.5% of AKI survivors. No infants were referred to a nephrologist for AKI follow-up.

Conclusions

AKI occurred in 40% of very low birth weight infants and was concentrated in the most premature and severely ill infants. One in six infants experienced multiple episodes of AKI, and a small number of infants was discharged with an elevated serum creatinine. Reporting a history of AKI in the discharge summary occurred infrequently, and referral to a nephrologist for AKI follow-up did not occur, highlighting areas for quality improvement.

Introduction

There is convincing evidence that the occurrence of AKI is associated with the development of CKD in both adult and pediatric patients (1–12). Although debate continues over whether this represents a unidirectional causal pathway, recent Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guidelines recommend that all patients who experience AKI be reassessed 3 months after AKI occurrence to evaluate for new onset or worsening of preexisting CKD (13). Patients without CKD at this initial follow-up are to be considered high risk for CKD development and managed accordingly (13,14).

Compared with other populations (15), data on the incidence of AKI in very low birth weight (VLBW; birth weight <1500 g) infants remain limited (16). Previous studies in neonates have typically used binary definitions of AKI on the basis of need for dialysis or a large increase in serum creatinine (typically to >1.5 mg/dl [>133 mmol/L]) (17). These studies underestimate the true incidence of AKI, because rises in serum creatinine as small as 0.3 mg/dl (26.5 mmol/L) are associated with increased morbidity and mortality (13). Recent studies have largely focused on specific subgroups (such as neonates with birth asphyxia [18,19], congenital diaphragmatic hernia [20], or congenital heart disease [21]), making it difficult to determine the risk for the broader population of VLBW infants.

Newer categorical AKI definitions on the basis of change in serum creatinine and duration of oliguria (such as the pediatric RIFLE [pRIFLE] [22], Acute Kidney Injury Network [AKIN] [23], and KDIGO [24] definitions) better classify the spectrum of AKI and allow consistent comparisons of AKI severity for research and clinical care. In an effort to adapt a standardized definition of AKI to neonates, Jetton and Askenazi (25) proposed a modified KDIGO definition on the basis of change in serum creatinine (Table 1). Although the National Institutes of Health have identified neonatal AKI as a high-priority area for research (26), there remain limited data applying this definition in VLBW infants (27). Moreover, there are no data on how frequently neonates with AKI have that history reported in the discharge summary—a prerequisite for appropriate follow-up—or are referred to a pediatric nephrologist at discharge.

Table 1.

Modified KDIGO definition of AKI used in the study

AKI Stage	Definition
0	No significant change in creatinine
1	↑ SCr by 0.3 mg/dl within 48 h or
	↑ in SCr by 150% to <200% from previous trough
2	↑ in SCr by 200% to <300% of previous trough
3	↑ in SCr≥300% of previous trough or
	SCr≥2.5 mg/dl or
	RRT

SCr, serum creatinine.

We, therefore, sought to evaluate the incidence of AKI in VLBW infants using the modified KDIGO definition, determine the association of AKI with clinically important outcomes, describe the frequency with which a diagnosis of AKI was reported in the discharge summary, and evaluate how frequently patients were referred to a pediatric nephrologist for AKI follow-up.

Materials and Methods

Study Population.

We reviewed all VLBW infants admitted to the University of Virginia’s level IV neonatal intensive care unit (NICU) from January 1, 2008, to December 31, 2011. All infants with birth weight ≤1500 g were initially included. Infants who survived for <2 days were excluded, because there were often insufficient numbers of serum creatinine measurements to classify these patients as having AKI. Additionally, patients admitted at >2 days of age were excluded, because laboratory data before transfer were often inaccessible. The Institutional Review Board at the University of Virginia approved the study protocol and waived the need for consent.

Data Collection and Definitions.

Data were collected from the electronic medical record (EMR) for each patient from admission until hospital discharge and included demographics, gestational age, birth weight, Apgar scores, laboratory results, length of stay, mortality, need for dialysis, and follow-up appointments made at the time of NICU discharge. Infants were classified as small, appropriate or large for gestational age using previously reported reference standards (28). Illness severity at admission was assessed using the updated clinical risk index for babies (CRIB II) score (29).

AKI was classified using the modified KDIGO definition previously proposed for use in neonates (25) (Table 1). For patients with multiple episodes of AKI, the highest stage reached for any episode was used for analysis. Serum creatinine was measured using an alkaline picrate (Jaffe) method traceable to isotope dilution mass spectrometry.

Any diagnosis, description, or statement that would have alerted a practitioner reading the discharge summary that the infant had experienced AKI was classified as a discharge summary that reported AKI, regardless of whether the term AKI was used. For example, infants whose discharge summaries included acute renal failure, renal insufficiency, azotemia, oliguria, or elevation in creatinine were all classified as having reported AKI in the discharge summary. Distinctions were not made between reporting in the final diagnosis list and describing the event within the body of the discharge summary. Discharge summaries that did not include any statement describing AKI were classified as having not reported AKI in the discharge summary.

The last measured creatinine for each infant before discharge was recorded. Values ≥0.5 mg/dl (44.2 mmol/L) were considered elevated if they were obtained at ≥4 weeks of life and ≥37 weeks postmenstrual age.

Statistical Analyses.

Continuous variables were compared using Mann–Whitney U or t tests, and categorical variables were compared using chi-squared or exact tests as appropriate. Linear and logistic regression models were created to determine the association between factors and outcomes of interest. Candidate variables for multivariable regression were identified on the basis of biologic plausibility or a P value <0.20 in univariable analysis. Multivariable models were constructed using backward variable selection among all candidate variables. Discrimination was determined by calculating receiver operating characteristic curves and their associated area under the curve. A two-sided significance level of 0.05 was set for all tests. All statistical analyses were performed using IBM SPSS, version 22 (Armonk, NY).

Results

In total, 535 VLBW infants were admitted to the NICU from January 1, 2008, to December 31, 2011. Of these, 455 infants were included in the AKI analyses, whereas 420 infants survived to discharge and were included in the survivor analyses. Patient identification, exclusion, and analyses are shown in Figure 1.

Figure 1.

Identification, exclusion, and analysis of infants for the study.

AKI Analyses

AKI occurred in 181 of 455 (39.8%) infants. Patient demographics and differences between infants who did and did not experience AKI are shown in Table 2. Most of these infants (58.6%) experienced only a single episode of AKI, but 75 (41.4%) infants had multiple episodes of AKI, with one infant experiencing seven discrete episodes of AKI.

Table 2.

Patient demographics with mean testing for infants who did and did not experience AKI

Characteristic	All Patients (n=455)		AKI (n=181)		No AKI (n=274)		P
	Mean (SD)	Range	Mean (SD)	Range	Mean (SD)	Range
Gestational age (wk)	27.7 (3)	22–37	25.7 (2.4)	22–33	29.0 (2.6)	24–37	<0.001
Birth weight (g)	1016 (296)	370–1495	816 (253)	370–1470	1148 (245)	400–1495	<0.001
Sex (% boys)	51		54		49		0.33
Apgar (1 min)	4.9 (2.7)	0–9	4.0 (2.6)	0–9	5.4 (2.5)	0–9	<0.001
Apgar (5 min)	6.9 (2.0)	0–9	6.0 (2.2)	0–9	7.4 (1.7)	2–9	<0.001
SGA (%)	22		14		26		0.002
CRIB II scorea	9.5 (3.9)	2–19	12.1 (3.6)	3–19	7.6 (3.0)	2–17	<0.001
Length of stay (d)	68 (41)	3–307	95 (69–114)b	3–307	46 (31–69)b	3–175	<0.001
Initial creatinine (mg/dl)	0.74 (0.14)	0.4–1.5	0.76 (0.15)	0.4–1.1	0.72 (0.13)	0.5–1.5	0.001
Mortality (%)	8		14		3		<0.001

SGA, small for gestational age; CRIB II, updated clinical risk index for babies score.

^aCRIB II scores were missing in 18.5% of patients. For all patients, n=371. For patients with AKI, n=157. For patients with no AKI, n=214.

^bAll results are reported as mean and SD except length of stay, which is reported as median and interquartile range.

When classified using the modified KDIGO definition, 117 (25.7%) infants reached stage 1, 51 (11.2%) infants reached stage 2, and 13 (2.9%) infants reached stage 3. One patient received peritoneal dialysis.

AKI was associated with lower gestational age and birth weight in univariate and multivariate logistic regression models (Supplemental Table 1). All 29 infants with gestational age <24 weeks experienced AKI versus none of the infants born at ≥34 weeks gestational age (Figure 2).

Figure 2.

AKI occurrence by gestational age among very low birth weight infants; 80% of infants who experienced AKI were born at <28 weeks gestational age, whereas 73% of infants who did not experience AKI were born at ≥28 weeks gestational age (odds ratio, 10.6; 95% confidence interval, 6.8 to 16.7; c statistic=0.76; 95% confidence interval, 0.72 to 0.81).

The CRIB II score was calculated for 371 (81.5%) patients. Greater illness severity as assessed by the CRIB II score was associated with more frequent AKI. A CRIB II score >10 was particularly associated with AKI: 109 of 141 (77.3%) patients with CRIB II>10 experienced AKI versus 48 of 230 (20.9%) patients with CRIB II≤10 (odds ratio [OR], 12.9; 95% confidence interval [95% CI], 7.8 to 21.4). As a predictor of AKI, a CRIB II score >10 yielded a c statistic of 0.78 (95% CI, 0.72 to 0.82).

AKI occurred in the first week of life in 116 of 181 (64.1%) infants (median=5 days; interquartile range [IQR]=3–11 days). The median age at first AKI was higher among infants ≥28 weeks gestational age than those <28 weeks gestational age (9 versus 5 days; P=0.001). Infants who experienced AKI within the first week of life had lower initial creatinine than those who did not (0.70 versus 0.76 mg/dl; P<0.001). There was no difference between the initial creatinine for infants who never experienced AKI and those who experienced AKI only after the first week of life.

Thirty-five (8%) infants died before NICU discharge. AKI was associated with mortality: 14.4% of infants with AKI died versus 3.3% of those without AKI (OR, 4.94; 95% CI, 2.26 to 10.81). The association between AKI and mortality persisted after accounting for confounding variables (Table 3, Supplemental Table 2). When stratified by stage, patients with stages 2 or 3 AKI were more likely to experience mortality than those with stage 1 AKI (Supplemental Table 3). There was no association between stage 1 AKI and mortality in multivariable models after adjusting for gestational age, birth weight, and CRIB II score.

Table 3.

Uni- and multivariable logistic regression analyses for mortality

Factor	OR	95% CI	P
Univariable analyses
AKI	4.94	2.26 to 10.81	<0.001
Apgar (1 min)	0.83	0.72 to 0.96	0.01
Apgar (5 min)	0.74	0.64 to 0.87	<0.001
Birth weight (100 g)	0.81	0.72 to 0.92	0.001
CRIB II	1.17	1.06 to 1.29	0.001
Gestational age (wk)	0.80	0.70 to 0.92	0.001
Preeclampsia	0.60	0.18 to 2.01	0.41
Race (non-Caucasian)	1.17	0.58 to 2.35	0.66
Sex (boys)	1.71	0.84 to 3.48	0.14
SGA	0.90	0.38 to 2.14	0.82
Multivariable analyses (final model)
AKI	4.00	1.39 to 11.52	0.01
Birth weight (100 g)	1.39	1.01 to 1.91	0.04
CRIB II	1.34	1.06 to 1.69	0.02
SGA	4.38	1.39 to 13.80	0.01

OR, odds ratio; 95% CI, 95% confidence interval; CRIB II, updated clinical risk index for babies; SGA, small for gestational age.

Serum creatinine measurements were obtained a median of 20 times per patient before death or discharge (IQR=9–38). After adjusting for length of stay, infants had serum creatinine measurements obtained a median of 0.35 times per day (IQR=0.23–0.48).

Survivor Analyses

Among 420 infants who survived to discharge, 36.9% experienced AKI. Of these infants, 13.5% had AKI reported in their discharge summary. AKI reporting was dependent on maximum AKI stage attained (P<0.001). Only 7.5% of infants with stage 1 AKI had AKI reported versus 21.4% and 57.1% of infants with stages 2 and 3 AKI, respectively. In a binary logistic regression model, every 0.1-mg/dl higher peak creatinine was associated with higher odds of AKI reporting by a factor of 1.45 (95% CI, 1.24 to 1.70; P<0.001). The number of episodes of AKI was also associated with AKI reporting (OR, 2.05; 95% CI, 1.41 to 2.99; P<0.001). There was no association between length of stay and AKI reporting.

AKI occurrence was associated with longer hospital stay. In a multivariable linear regression model (Table 4), AKI occurrence was associated with a 12-day longer stay (95% CI, 5.1 to 18.4) after adjusting for birth weight, gestational age, and CRIB II score.

Table 4.

Uni- and multivariable linear regression for length of stay (days)

Factor	Days	95% CI	P
Univariable analyses
AKI (occurrence)	53.3	49.5 to 57.1	<0.001
AKI (stage 1 only)	39.4	32.7 to 46.1	<0.001
AKI (no. of episodes)	19.1	16.3 to 22.0	<0.001
Apgar (1 min)	−3.6	−4.9 to −2.3	<0.001
Apgar (5 min)	−5.4	−7.1 to −3.6	<0.001
Birth weight (100 g)	−9.2	−10.1 to −8.4	<0.001
CRIB II	11.2	3.8 to 18.7	0.003
Gestational age (wk)	−9.4	−10.3 to −8.6	<0.001
Preeclampsia	−2.6	−12.8 to 7.5	0.61
Race (non-Caucasian)	−1.7	−8.9 to 5.5	0.64
Sex (boys)	3.3	−3.7 to 10.4	0.36
SGA	−24.9	−32.9 to −16.9	<0.001
Multivariable analyses (final models)
AKI (occurrence)	11.7	5.1 to 18.4	0.001
Birth weight (100 g)	−2.9	−4.8 to −1.1	0.002
CRIB II	1.1	−0.7 to 3.0	0.23
Gestational age (wk)	−5.1	−7.6 to −2.6	<0.001
AKI (stage 1 only)	5.4	−0.8 to 11.5	0.09
Birth weight (100 g)	−2.0	−3.6 to −0.3	0.02
CRIB II	1.9	0.3 to 3.6	0.02
Gestational age (wk)	−5.7	−7.9 to −3.6	<0.001
AKI (no. of episodes)	4.9	2.0 to 7.7	0.001
Birth weight (100 g)	−3.0	−4.8 to −1.1	0.002
CRIB II	1.4	−0.5 to 3.2	0.15
Gestational age (wk)	−4.8	−7.3 to −2.3	<0.001

CRIB II, updated clinical risk index for babies; SGA, small for gestational age.

In 2010, an EMR was implemented. The EMR captures diagnoses made during the hospital stay and lists them on the discharge summary, eliminating the need to retrospectively identify AKI. There was a nonsignificant trend toward improved AKI reporting after implementation of the EMR (pre-EMR, 10%; post-EMR, 20%; P=0.08).

Among 155 infants who experienced AKI and survived to discharge, 3.9% of infants were referred to a pediatric nephrologist, similar to the referral rate among infants who did not experience AKI (3.0%; P=0.64). No infant was referred specifically for follow-up of AKI. The 14 infants who were referred to a pediatric nephrologist had the following diagnoses: hypertension (three infants), hydronephrosis (three infants), metabolic acidosis (two infants), nephrocalcinosis (one infant), pelvic kidney (two infants), glycosuria (one infant), and vesicoureteral reflux (two infants).

There were 126 infants whose last recorded creatinine was obtained at ≥4 weeks of age and ≥37 weeks postmenstrual age. Eight of these infants had a last recorded creatinine of 0.5 mg/dl (44.2 mmol/L) or higher. All of these infants had experienced AKI (P=0.02); among these, one infant was referred to a nephrologist for hypertension.

Discussion

In this large retrospective study of VLBW infants, AKI was common and occurred more frequently than previously reported (27,30). Using a modified KDIGO definition, AKI occurred in nearly 40% of all VLBW infants and 37% of NICU survivors. Similar to previous reports, AKI was associated with mortality and increased length of stay, and AKI occurrence was concentrated among the smallest and most premature infants. However, despite the frequency with which AKI occurred, communicating that diagnosis in the discharge summary occurred for only a minority of patients.

Premature and low birth weight infants have a greater long-term risk of developing CKD than the general population (9,31,32). The contribution of AKI to this risk is unclear. Nephrogenesis is not complete until 36 weeks of gestation, meaning that infants born prematurely must continue to form nephrons ex utero, where that process may be abbreviated (33) or abnormal (34). Although it was previously thought that most AKI was reversible, experimental data indicate that recovery from AKI is incomplete, with intrinsic forms of AKI causing irreversible damage (35–37). It may, therefore, be cause for concern that nearly two thirds of VLBW infants who experienced AKI did so during the first week of life.

It is possible that AKI is not an additive risk factor for CKD development and that its occurrence may simply identify a subpopulation already at risk for CKD because of low nephron number or colinearity with other risk factors for CKD (38). However, in either case, the implications for follow-up are the same, and the occurrence of AKI may be an opportunity to identify patients most at risk for developing CKD (39). It is noteworthy that 2% of infants in our study were discharged with an elevated last creatinine. All of these infants had experienced AKI, suggesting that some infants who experienced AKI could have had CKD even at the time of discharge.

Modern AKI classifications—such as KDIGO and AKIN—are intended to be more sensitive than previous definitions. Our findings suggest that this finding is the case. In 2012, Viswanathan et al. (30) reviewed 472 extremely low birth weight (ELBW) infants and found that 12.5% experienced AKI using a definition of “oliguria < 1 ml/kg/h of urine output that developed 24 hours after birth and persisted for at least 24 hours and/or a raising of the serum creatinine level >1.5 mg/dl 72 hours after birth in the presence of normal maternal creatinine levels” (30). Although we did not assess oliguria, 11.7% of 205 ELBW infants in our population experienced AKI when defined as a creatinine>1.5 mg/dl (>133 mmol/L). However, AKI defined by the modified KDIGO classification occurred in 69% of infants, with 94% of these infants experiencing stages 1 or 2 AKI.

Although mild AKI may seem insignificant, some pediatric patients with only mild or moderate AKI develop CKD or risk factors for it (such as hypertension or hyperfiltration) when followed long term (1,7), and even small increases in creatinine are associated with short-term risk. A rise in creatinine of 0.2–0.3 mg/dl independently predicts mortality in adults (40), and similarly, small rises in pediatric patients (25%–50% of baseline) are associated with increased length of stay and ventilator days in children undergoing cardiac surgery (41). Within our population, AKI was associated with higher mortality and a 12-day increase in length of stay, even after accounting for confounding factors.

Our results highlight some of the potential issues with creatinine-based classification systems. Neonates begin life with an elevated creatinine, which then drops over the first weeks of life (42). Percentage-based AKI definitions (such as pRIFLE, AKIN, and KDIGO definitions) are unable to capture neonates whose creatinine does not drop as expected after birth. Additionally, an initially elevated creatinine is protective against being classified as having AKI: in our study, infants who experienced AKI within the first week of life had lower initial creatinine than those who did not (0.70 versus 0.76 mg/dl; P<0.001). Similar findings have been reported among neonates undergoing cardiac surgery using the AKIN criteria (43).

Nonetheless, our data suggest that, if mild AKI is of any long-term consequence, its occurrence is poorly communicated to other providers in the discharge summary. In our population, only 13.5% of patients who experienced AKI were identified as such in their discharge summary. The reasons for such poor reporting are multifactorial. In the final years of our study, our institution began using an EMR that automatically incorporated inpatient diagnoses into the discharge summary. Rates of AKI reporting remained low even with the EMR, suggesting that the problem with AKI reporting is as much one of recognition as it is one of recall. Although electronic technology has the potential to improve discharge handoffs, efforts to improve AKI recognition are needed as well.

Our study has several limitations. First, in this retrospective review, we were unable to reliably determine the cause of AKI or fully assess confounders, making it difficult to determine why AKI occurred more frequently than has been reported in studies using similar classification systems (27). Although we did not quantify exposure to nephrotoxins in all infants, we have previously described this within a subset of this cohort, finding exposure to gentamicin in 87% of infants, indomethacin in 43% of infants, and vancomycin in 25% of infants (44). Nephrotoxic medications (particularly nonsteroidal anti-inflammatory drugs) are a well described cause of decreased renal function in neonates (45,46) and could explain the high incidence of AKI. It is also possible that the higher incidence of AKI reflects more frequent monitoring of serum creatinine at our institution, because each infant had, on average, one serum creatinine measurement obtained for every 2.8 hospital days.

Second, we did not assess oliguria, which increases the sensitivity of AKI classification and might have identified additional patients with AKI. However, the retrospective design may reduce selection bias, because we investigated all patients who met inclusion criteria.

Third, our laboratory uses an alkaline picrate (Jaffe) method to measure creatinine. This technique is subject to interference from albumin, immunoglobulins, unconjugated bilirubin, and hemoglobins (both A and F) (47), and it may overestimate serum creatinine by 0.2 mg/dl in ELBW infants <6 weeks old (48). Because the KDIGO AKI classification relies on percentage change in creatinine, this might have led to an underestimate of AKI prevalence.

Fourth, we classified a last creatinine measurement ≥0.5 mg/dl (44.2 mmol/L) at ≥4 weeks of age and ≥37 weeks postmenstrual age as abnormal and potentially indicative of CKD. Although such values are ≥20% higher than the 97.5th percentile for healthy term infants (49), this threshold may not appropriately classify preterm neonates. Data are lacking regarding what constitutes normal serum creatinine in extremely preterm infants at varying postnatal ages; without prospective follow-up, it is unclear whether previously published ranges (50) are truly normal (i.e., reflective of good health) or simply normative (i.e., reflecting a baseline incidence of CKD). Additionally, the circumstances under which the last creatinine was obtained are not clear retrospectively and might not indicate the infant’s steady-state renal function.

In conclusion, in a large single-center review of VLBW infants, AKI was common (39.8%) and concentrated in the smallest, most premature, and severely ill infants, particularly those born at <28 weeks gestational age or who had a CRIB II score >10. Although most AKI was mild, 16.5% of infants experienced multiple episodes of AKI, and a small number of infants were discharged with an elevated serum creatinine. Reporting an infant’s history of AKI at discharge occurred infrequently, and no infant was referred to a pediatric nephrologist for AKI follow-up. These findings represent significant areas for quality improvement.

Disclosures

None.